Enhancing Job Accessibility and Opportunities in Melbourne

Authored by: Tharaka Sandamal And Sahan Chamod

Duration: 180 mins

Level: Intermediate

Pre-requisite Skills: Python, Data Analysis, Pandas, Data Visualisation

Scenario

Use Case Summary: Job Accessibility Analysis in Melbourne

Scenario

Melbourne displays significant variation in spatial job accessibility, which impacts economic opportunities and social inclusion. Effective planning requires a holistic understanding of how employment locations, public transport availability, and commercial infrastructure interact across the city. However, current decision-making often lacks unified spatial analysis, making it challenging to strategically address transport gaps in high-demand employment areas.

---

Project Objective

The objective of this project is to evaluate and improve job accessibility in Melbourne by integrating employment density, transport stop coverage, and commercial floor space. The project focuses on identifying areas where transport access is insufficient relative to job and business activity, using both rule-based and machine learning-driven approaches.

---

Project Stages

Stage 1: Data Collection & Integration

• Collect datasets from:

  • Employment by Block by CLUE Industry (2023)
  • Bus and Tram Stops from City of Melbourne open data
  • Floor Space by Use by Block (2023)
  • CLUE Block Boundaries (GeoJSON format)

• Preprocess data by validating geometries, aligning coordinate reference systems, handling missing values, and performing spatial joins to merge datasets into a unified format for analysis.

Stage 2: Exploratory Data Analysis (EDA)

• Analyse employment distribution across city blocks and CLUE small areas.
• Explore floor space patterns to understand commercial concentration and potential business zones.
• Visualise spatial distribution of bus and tram stops using interactive maps with marker clustering.
• Aggregate transport stop counts per block and compute basic ratios to assess coverage.

Stage 3: Spatial Accessibility Analysis

• Compute transport-to-worker and transport-to-space ratios to quantify accessibility at the block level.
• Visualise accessibility metrics using choropleth maps and interactive GeoJSON-based mapping.
• Apply rule-based filtering to identify blocks with high demand (jobs or space) but low accessibility ratios.
• Integrate machine learning clustering (KMeans) to classify blocks into accessibility levels: Low, Medium, and High.
• Generate an interactive final map displaying cluster labels, block information, and transport overlays to communicate accessibility insights clearly.

Stage 4: Visualisation and Dashboard Development

• Develop interactive folium-based maps with informative tooltips showing block ID, CLUE area, transport stops, worker and space ratios, and cluster level.
• Use marker clusters to display bus and tram stop locations for layered analysis.
• Implement a customised legend and colour scheme to represent accessibility tiers clearly.
• Ensure maps are interactive and suitable for stakeholder presentations and urban planning use.

Stage 5: Recommendations and Strategic Planning

• Identify blocks in Low Accessibility (Cluster 0) that show high commercial floor space or employment potential.
• Recommend prioritisation of transport infrastructure enhancements in these blocks.
• Highlight blocks with high accessibility but under-utilised space as candidates for business development and employment programs.
• Suggest expansion by incorporating train network data and real-time flows for future analysis stages.

---

Expected Outcomes

• Deliver a complete spatial model highlighting job accessibility gaps.
• Enable data-driven decisions for infrastructure planning and economic development.
• Guide resource allocation by pinpointing priority blocks for intervention.
• Present an expandable methodology that supports ongoing monitoring and refinement of accessibility planning.

---

1. Employment by Block by CLUE Industry (2002–2023)

• Contains employment data showing job counts per city block, categorised by industry and small area allocation.
• Valuable for analysing spatial job density, sectoral employment patterns, and long-term trends.
• Link: Employment by Block by CLUE Industry

---

2. Metro Train Stations with Accessibility Information

• Details the locations of metro train stations along with accessibility features.
• Crucial for evaluating public transport coverage and accessibility to employment areas.
• Link: Metro Train Stations with Accessibility Information

---

3. Bus Stops

• Provides geospatial locations of bus stops across Melbourne.
• Important for assessing bus network coverage and accessibility gaps relative to employment opportunities.
• Link: Bus Stops

---

4. Floor Space by Use by Block (2002–2023)

• Offers data on floor space usage categorised by type (commercial, office, retail, etc.) and city block.
• Useful for determining business concentration, commercial capacity, and potential for employment growth.
• Link: Floor Space by Use by Block

---

5. City Circle Tram Stops

• Provides geospatial data on the stops of the City Circle tram service operating within Melbourne's central business district.
• Essential for analysing tram network coverage, especially in the CBD, and assessing accessibility to key employment and tourist areas.
• Link: City Circle Tram Stops

Package/Library Imports

IMPORT LIBRARIES

# General purpose libraries
import numpy as np
import pandas as pd
from io import StringIO
import requests
import json

# Configuration
from config import API_KEY  # Securely import API key

# Geospatial processing
import geopandas as gpd
from shapely.geometry import Point, shape

# Visualisation
import matplotlib.pyplot as plt
import folium
from folium import Choropleth
from folium.plugins import MarkerCluster
from folium.features import GeoJson, GeoJsonTooltip

# Machine learning
from sklearn.preprocessing import StandardScaler
from sklearn.cluster import KMeans

Read data using API

#from config import API_KEY  # Import API key securely


# Preferred Method: Export Endpoint

# Function to collect data from the City of Melbourne Open Data API
def collect_data(dataset_id):
    base_url = 'https://data.melbourne.vic.gov.au/api/explore/v2.1/catalog/datasets/'
    format = 'csv'
    url = f'{base_url}{dataset_id}/exports/{format}'
    
    params = {
        'select': '*',
        'limit': -1,  # all records
        'lang': 'en',
        'timezone': 'UTC'
        #'api_key': API_KEY  # use if datasets require API key permissions
    }

    # GET request
    response = requests.get(url, params=params)
    if response.status_code == 200:
        url_content = response.content.decode('utf-8')
        dataset = pd.read_csv(StringIO(url_content), delimiter=';')
        return dataset 
    else:
        print(f'Request failed with status code {response.status_code}')

# Read data using the function
#transport_df = collect_data('metro-train-stations-with-accessibility-information')
emp_block_df = collect_data('employment-by-block-by-clue-industry')
block_df = collect_data('blocks-for-census-of-land-use-and-employment-clue')
Bus_df = collect_data('bus-stops')
floor_df = collect_data('floor-space-by-use-by-block')
tram_df = collect_data('city-circle-tram-stops')
clue_areas_df = collect_data('small-areas-for-census-of-land-use-and-employment-clue')

Step 1.2: Data Inspection

Employment by Block by CLUE Industry

# Preview the dataset
print("First 5 records:")
display(emp_block_df.head())

# Check shape and column names
print("Shape:", emp_block_df.shape)
print("\nColumn Names:")
print(emp_block_df.columns.tolist())

# Check data types
print("\nData Types:")
print(emp_block_df.dtypes)

# Check for missing values
print("\nMissing Values:")
print(emp_block_df.isnull().sum())


# Get unique industries if present
if 'industry' in emp_block_df.columns:
    print("\nUnique Industries:")
    print(emp_block_df['industry'].unique())

First 5 records:

	census_year	block_id	clue_small_area	accommodation	admin_and_support_services	agriculture_and_mining	arts_and_recreation_services	business_services	construction	education_and_training	...	information_media_and_telecommunications	manufacturing	other_services	public_administration_and_safety	real_estate_services	rental_and_hiring_services	retail_trade	transport_postal_and_storage	wholesale_trade	total_jobs_in_block
0	2023	6	Melbourne (CBD)	0.0	0.0	0.0	203.0	0.0	0.0	0.0	...	NaN	0.0	NaN	0.0	NaN	0.0	47.0	0.0	0.0	647.0
1	2023	11	Melbourne (CBD)	187.0	40.0	0.0	NaN	182.0	NaN	0.0	...	NaN	0.0	36.0	0.0	42.0	0.0	14.0	NaN	0.0	759.0
2	2023	12	Melbourne (CBD)	32.0	NaN	0.0	NaN	1525.0	0.0	379.0	...	NaN	0.0	21.0	NaN	NaN	0.0	11.0	NaN	0.0	4087.0
3	2023	14	Melbourne (CBD)	124.0	167.0	NaN	15.0	550.0	NaN	292.0	...	19.0	NaN	165.0	23.0	53.0	0.0	44.0	NaN	33.0	2049.0
4	2023	15	Melbourne (CBD)	75.0	103.0	NaN	149.0	120.0	NaN	253.0	...	73.0	24.0	220.0	NaN	NaN	0.0	266.0	NaN	NaN	1833.0

5 rows × 24 columns

Shape: (13000, 24)

Column Names:
['census_year', 'block_id', 'clue_small_area', 'accommodation', 'admin_and_support_services', 'agriculture_and_mining', 'arts_and_recreation_services', 'business_services', 'construction', 'education_and_training', 'electricity_gas_water_and_waste_services', 'finance_and_insurance', 'food_and_beverage_services', 'health_care_and_social_assistance', 'information_media_and_telecommunications', 'manufacturing', 'other_services', 'public_administration_and_safety', 'real_estate_services', 'rental_and_hiring_services', 'retail_trade', 'transport_postal_and_storage', 'wholesale_trade', 'total_jobs_in_block']

Data Types:
census_year                                   int64
block_id                                      int64
clue_small_area                              object
accommodation                               float64
admin_and_support_services                  float64
agriculture_and_mining                      float64
arts_and_recreation_services                float64
business_services                           float64
construction                                float64
education_and_training                      float64
electricity_gas_water_and_waste_services    float64
finance_and_insurance                       float64
food_and_beverage_services                  float64
health_care_and_social_assistance           float64
information_media_and_telecommunications    float64
manufacturing                               float64
other_services                              float64
public_administration_and_safety            float64
real_estate_services                        float64
rental_and_hiring_services                  float64
retail_trade                                float64
transport_postal_and_storage                float64
wholesale_trade                             float64
total_jobs_in_block                         float64
dtype: object

Missing Values:
census_year                                    0
block_id                                       0
clue_small_area                                0
accommodation                               2533
admin_and_support_services                  2103
agriculture_and_mining                       622
arts_and_recreation_services                4129
business_services                           1975
construction                                2173
education_and_training                      2668
electricity_gas_water_and_waste_services    2583
finance_and_insurance                       1650
food_and_beverage_services                  2755
health_care_and_social_assistance           2700
information_media_and_telecommunications    2386
manufacturing                               2417
other_services                              3433
public_administration_and_safety            1924
real_estate_services                        1888
rental_and_hiring_services                   890
retail_trade                                2634
transport_postal_and_storage                2340
wholesale_trade                             2618
total_jobs_in_block                         2821
dtype: int64

print("First 5 records:")
display(block_df.head())

First 5 records:

	geo_point_2d	geo_shape	block_id	clue_area
0	-37.788517214796904, 144.9421522923356	{"coordinates": [[[144.9421360326, -37.7875659...	925	Parkville
1	-37.78752314475404, 144.9428068448982	{"coordinates": [[[144.9421360326, -37.7875659...	924	Parkville
2	-37.78035632811396, 144.94536433435934	{"coordinates": [[[144.9425926939, -37.7787229...	930	Parkville
3	-37.80720183148242, 144.94336441344424	{"coordinates": [[[144.9427148261, -37.8073238...	412	West Melbourne (Residential)
4	-37.806706331914484, 144.94240782994808	{"coordinates": [[[144.9427148261, -37.8073238...	410	West Melbourne (Residential)

City Circle tram stops

tram_df.head()

# Preview City Circle Tram Stops dataset
print("First 5 records:")
display(tram_df.head())

# Check shape and column names
print("Shape:", tram_df.shape)

print("\nColumn Names:")
print(tram_df.columns.tolist())

# Data types
print("\nData Types:")
print(tram_df.dtypes)

# Missing values
print("\nMissing Values:")
print(tram_df.isnull().sum())

First 5 records:

	geo_point_2d	geo_shape	name	xorg	stop_no	mccid_str	xsource	xdate	mccid_int
0	-37.82023778673241, 144.95786314283018	{"coordinates": [144.95786314283018, -37.82023...	Melbourne Aquarium / Flinders Street	GIS Team	2	NaN	Mapbase	2011-10-18	4
1	-37.82097269970027, 144.95546153614245	{"coordinates": [144.95546153614245, -37.82097...	Spencer Street / Flinders Street	GIS Team	1	NaN	Mapbase	2011-10-18	5
2	-37.82190465062153, 144.95109855638137	{"coordinates": [144.95109855638137, -37.82190...	The Goods Shed / Wurundjeri Way	GIS Team	D5	NaN	Mapbase	2011-10-18	7
3	-37.811771476718356, 144.95644059700524	{"coordinates": [144.95644059700524, -37.81177...	William Street / La Trobe Street	GIS Team	3	NaN	Mapbase	2011-10-18	16
4	-37.81105928060848, 144.95891745116262	{"coordinates": [144.95891745116262, -37.81105...	Queen Street / La Trobe Street	GIS Team	4	NaN	Mapbase	2011-10-18	17

Shape: (28, 9)

Column Names:
['geo_point_2d', 'geo_shape', 'name', 'xorg', 'stop_no', 'mccid_str', 'xsource', 'xdate', 'mccid_int']

Data Types:
geo_point_2d     object
geo_shape        object
name             object
xorg             object
stop_no          object
mccid_str       float64
xsource          object
xdate            object
mccid_int         int64
dtype: object

Missing Values:
geo_point_2d     0
geo_shape        0
name             0
xorg             0
stop_no          0
mccid_str       28
xsource          0
xdate            0
mccid_int        0
dtype: int64

Bus Stops

# Preview Bus Stops dataset
print("First 5 records:")
display(Bus_df.head())

# Check shape and column names
print("Shape:", Bus_df.shape)
print("\nColumn Names:")
print(Bus_df.columns.tolist())

# Data types
print("\nData Types:")
print(Bus_df.dtypes)

# Missing values
print("\nMissing Values:")
print(Bus_df.isnull().sum())

First 5 records:

	geo_point_2d	geo_shape	prop_id	addresspt1	addressp_1	asset_clas	asset_type	objectid	str_id	addresspt	asset_subt	model_desc	mcc_id	roadseg_id	descriptio	model_no
0	-37.81842481335676, 144.96889648633675	{"coordinates": [144.96889648633675, -37.81842...	573333	29.149053	288	Signage	Sign - Public Transport	749	1249454	606816	NaN	Sign - Public Transport 1 Panel	1249454	0	Sign - Public Transport 1 Panel Bus Stop Type 12	P.16
1	-37.81767597164006, 144.95888238475013	{"coordinates": [144.95888238475013, -37.81767...	0	10.537902	105	Signage	Sign - Public Transport	2098	1247042	507646	NaN	Sign - Public Transport 1 Panel	1247042	20118	Sign - Public Transport 1 Panel Bus Stop Type 8	P.16
2	-37.78188915501449, 144.95963193312105	{"coordinates": [144.95963193312105, -37.78188...	0	25.269643	212	Signage	Sign - Public Transport	2143	1252383	108510	NaN	Sign - Public Transport 1 Panel	1252383	22387	Sign - Public Transport 1 Panel Bus Stop Type 8	P.16
3	-37.793726533289785, 144.94716743007305	{"coordinates": [144.94716743007305, -37.79372...	0	44.230506	237	Signage	Sign - Public Transport	2627	1249788	100087	NaN	Sign - Public Transport 1 Panel	1249788	20919	Sign - Public Transport 1 Panel Bus Stop Type 8	P.16
4	-37.80286165064241, 144.92778487963457	{"coordinates": [144.92778487963457, -37.80286...	0	67.718553	360	Signage	Sign - Public Transport	3306	1235311	103002	NaN	Sign - Public Transport 1 Panel	1235311	21680	Sign - Public Transport 1 Panel Bus Stop Type 13	P.16

Shape: (309, 16)

Column Names:
['geo_point_2d', 'geo_shape', 'prop_id', 'addresspt1', 'addressp_1', 'asset_clas', 'asset_type', 'objectid', 'str_id', 'addresspt', 'asset_subt', 'model_desc', 'mcc_id', 'roadseg_id', 'descriptio', 'model_no']

Data Types:
geo_point_2d     object
geo_shape        object
prop_id           int64
addresspt1      float64
addressp_1        int64
asset_clas       object
asset_type       object
objectid          int64
str_id            int64
addresspt         int64
asset_subt      float64
model_desc       object
mcc_id            int64
roadseg_id        int64
descriptio       object
model_no         object
dtype: object

Missing Values:
geo_point_2d      0
geo_shape         0
prop_id           0
addresspt1        0
addressp_1        0
asset_clas        0
asset_type        0
objectid          0
str_id            0
addresspt         0
asset_subt      309
model_desc        0
mcc_id            0
roadseg_id        0
descriptio        0
model_no          0
dtype: int64

Floor Space by Use by Block

# Preview floor space dataset
print("First 5 records:")
display(floor_df.head())

# Check shape and columns
print("Shape:", floor_df.shape)
print("\nColumn Names:")
print(floor_df.columns.tolist())

# Data types
print("\nData Types:")
print(floor_df.dtypes)

# Missing values
print("\nMissing Values:")
print(floor_df.isnull().sum())

First 5 records:

	census_year	block_id	clue_small_area	commercial_accommodation	common_area	community_use	educational_research	entertainment_recreation_indoor	equipment_installation	hospital_clinic	...	transport	transport_storage_uncovered	unoccupied_under_construction	unoccupied_under_demolition_condemned	unoccupied_under_renovation	unoccupied_undeveloped_site	unoccupied_unused	wholesale	workshop_studio	total_floor_space_in_block
0	2007	527	Kensington	0.0	NaN	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	NaN	0.0	0.0	0.0	21138.0
1	2007	529	Kensington	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	NaN	0.0	0.0	0.0	10098.0
2	2007	534	Kensington	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	NaN
3	2007	539	Kensington	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	NaN	NaN	0.0	0.0	49600.0
4	2007	542	Kensington	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	1735.0

5 rows × 41 columns

Shape: (13000, 41)

Column Names:
['census_year', 'block_id', 'clue_small_area', 'commercial_accommodation', 'common_area', 'community_use', 'educational_research', 'entertainment_recreation_indoor', 'equipment_installation', 'hospital_clinic', 'house_townhouse', 'institutional_accommodation', 'manufacturing', 'office', 'park_reserve', 'parking_commercial_covered', 'parking_commercial_uncovered', 'parking_private_covered', 'parking_private_uncovered', 'performances_conferences_ceremonies', 'private_outdoor_space', 'public_display_area', 'residential_apartment', 'retail_cars', 'retail_shop', 'retail_showroom', 'retail_stall', 'sports_and_recreation_outdoor', 'square_promenade', 'storage', 'student_accommodation', 'transport', 'transport_storage_uncovered', 'unoccupied_under_construction', 'unoccupied_under_demolition_condemned', 'unoccupied_under_renovation', 'unoccupied_undeveloped_site', 'unoccupied_unused', 'wholesale', 'workshop_studio', 'total_floor_space_in_block']

Data Types:
census_year                                int64
block_id                                   int64
clue_small_area                           object
commercial_accommodation                 float64
common_area                              float64
community_use                            float64
educational_research                     float64
entertainment_recreation_indoor          float64
equipment_installation                   float64
hospital_clinic                          float64
house_townhouse                          float64
institutional_accommodation              float64
manufacturing                            float64
office                                   float64
park_reserve                             float64
parking_commercial_covered               float64
parking_commercial_uncovered             float64
parking_private_covered                  float64
parking_private_uncovered                float64
performances_conferences_ceremonies      float64
private_outdoor_space                    float64
public_display_area                      float64
residential_apartment                    float64
retail_cars                              float64
retail_shop                              float64
retail_showroom                          float64
retail_stall                             float64
sports_and_recreation_outdoor            float64
square_promenade                         float64
storage                                  float64
student_accommodation                    float64
transport                                float64
transport_storage_uncovered              float64
unoccupied_under_construction            float64
unoccupied_under_demolition_condemned    float64
unoccupied_under_renovation              float64
unoccupied_undeveloped_site              float64
unoccupied_unused                        float64
wholesale                                float64
workshop_studio                          float64
total_floor_space_in_block               float64
dtype: object

Missing Values:
census_year                                 0
block_id                                    0
clue_small_area                             0
commercial_accommodation                 2343
common_area                              3236
community_use                             925
educational_research                     2709
entertainment_recreation_indoor          2655
equipment_installation                   3288
hospital_clinic                          1866
house_townhouse                          1088
institutional_accommodation               527
manufacturing                            1000
office                                   2151
park_reserve                             2414
parking_commercial_covered               1935
parking_commercial_uncovered              536
parking_private_covered                  3111
parking_private_uncovered                1367
performances_conferences_ceremonies      2219
private_outdoor_space                    2687
public_display_area                       672
residential_apartment                    3212
retail_cars                               325
retail_shop                              2193
retail_showroom                          1117
retail_stall                              568
sports_and_recreation_outdoor             436
square_promenade                          245
storage                                  2828
student_accommodation                     956
transport                                 328
transport_storage_uncovered               204
unoccupied_under_construction            1296
unoccupied_under_demolition_condemned     331
unoccupied_under_renovation              1615
unoccupied_undeveloped_site              1776
unoccupied_unused                        2527
wholesale                                1063
workshop_studio                          2452
total_floor_space_in_block                620
dtype: int64

Preparing CLUE Block Dataset for Mapping

• The dataset includes polygon shapes for each CLUE block area.
• These polygons are converted into a GeoDataFrame for spatial visualisation and joining.

# Import necessary modules
#from shapely.geometry import shape
#import json

# Convert 'geo_shape' column into actual geometry
block_df['geometry'] = block_df['geo_shape'].apply(lambda x: shape(json.loads(x)))

# Convert to GeoDataFrame
block_gdf = gpd.GeoDataFrame(block_df, geometry='geometry')
block_gdf.set_crs(epsg=4326, inplace=True)

# Preview
print("Block Geodata - Remaining missing values:", block_gdf.isnull().sum().sum())
block_gdf.head()

Block Geodata - Remaining missing values: 0

	geo_point_2d	geo_shape	block_id	clue_area	geometry
0	-37.788517214796904, 144.9421522923356	{"coordinates": [[[144.9421360326, -37.7875659...	925	Parkville	POLYGON ((144.94214 -37.78757, 144.941 -37.788...
1	-37.78752314475404, 144.9428068448982	{"coordinates": [[[144.9421360326, -37.7875659...	924	Parkville	POLYGON ((144.94214 -37.78757, 144.94319 -37.7...
2	-37.78035632811396, 144.94536433435934	{"coordinates": [[[144.9425926939, -37.7787229...	930	Parkville	POLYGON ((144.94259 -37.77872, 144.94202 -37.7...
3	-37.80720183148242, 144.94336441344424	{"coordinates": [[[144.9427148261, -37.8073238...	412	West Melbourne (Residential)	POLYGON ((144.94271 -37.80732, 144.9437 -37.80...
4	-37.806706331914484, 144.94240782994808	{"coordinates": [[[144.9427148261, -37.8073238...	410	West Melbourne (Residential)	POLYGON ((144.94271 -37.80732, 144.94285 -37.8...

Step 1.3: Data Cleaning

Employment Dataset – Handle Missing Values and Recalculate Totals

• Missing values in job sector columns are assumed to represent zero jobs.
• Fill them with 0 and recalculate the total number of jobs in each block.

# Fill missing job sector values with 0
emp_block_df.fillna(0, inplace=True)

# Recalculate total jobs in block from individual sectors
sector_columns = emp_block_df.columns[3:-1]
emp_block_df['total_jobs_in_block'] = emp_block_df[sector_columns].sum(axis=1)

# Confirm all missing values are handled
print("Employment Data - Remaining missing values:", emp_block_df.isnull().sum().sum())

Employment Data - Remaining missing values: 0

Cleaning Metro Train Stations Dataset

• Split the geo_point_2d column into latitude and longitude.
• Convert the dataset into a GeoDataFrame for spatial analysis and mapping.

#from shapely.geometry import Point
#import geopandas as gpd

# Extract latitude and longitude
tram_df[['latitude', 'longitude']] = tram_df['geo_point_2d'].str.split(',', expand=True).astype(float)

# Convert into GeoDataFrame
tram_geometry = [Point(xy) for xy in zip(tram_df['longitude'], tram_df['latitude'])]
tram_gdf = gpd.GeoDataFrame(tram_df, geometry=tram_geometry)
tram_gdf.set_crs(epsg=4326, inplace=True)

# Check missing values
print("City Circle Tram Stops Data - Remaining missing values:", tram_gdf.isnull().sum().sum())

City Circle Tram Stops Data - Remaining missing values: 28

• Only mccid_str has missing values in all records.
• Since mccid_int is complete and likely sufficient, mccid_str is safely dropped to avoid redundancy.

# Drop the redundant column with all missing values
tram_df.drop(columns=['mccid_str'], inplace=True)

# Confirm remaining missing values
print("City Circle Tram Stops Data – Remaining missing values:", tram_df.isnull().sum().sum())

City Circle Tram Stops Data – Remaining missing values: 0

Cleaning Bus Stops Dataset

• Coordinates are split from the geo_point_2d field.
• Data is converted into a GeoDataFrame with EPSG:4326 CRS for mapping.

# Extract latitude and longitude
Bus_df[['latitude', 'longitude']] = Bus_df['geo_point_2d'].str.split(',', expand=True).astype(float)

# Create GeoDataFrame
bus_geometry = [Point(xy) for xy in zip(Bus_df['longitude'], Bus_df['latitude'])]
bus_gdf = gpd.GeoDataFrame(Bus_df, geometry=bus_geometry)
bus_gdf.set_crs(epsg=4326, inplace=True)

# Preview
bus_gdf.head()

print("Bus Stops Data - Remaining missing values:", bus_gdf.isnull().sum().sum())

Bus Stops Data - Remaining missing values: 309

Check Column-Wise Missing Value Count

# Show missing values per column
bus_gdf.isnull().sum()

geo_point_2d      0
geo_shape         0
prop_id           0
addresspt1        0
addressp_1        0
asset_clas        0
asset_type        0
objectid          0
str_id            0
addresspt         0
asset_subt      309
model_desc        0
mcc_id            0
roadseg_id        0
descriptio        0
model_no          0
latitude          0
longitude         0
geometry          0
dtype: int64

The asset_subt column is completely missing for all records. Since it appears to be a numeric field and is not essential for the spatial or analytical objectives of the project, missing values are replaced with 0.

# Replace missing numeric values in 'asset_subt' with 0 safely
bus_gdf['asset_subt'] = bus_gdf['asset_subt'].fillna(0)

# Confirm all missing values are handled
print("Bus Stops Data – Remaining missing values:", bus_gdf.isnull().sum().sum())

Bus Stops Data – Remaining missing values: 0

Cleaning Floor Space by Use by Block

• Most missing values likely represent zero allocated floor space.
• Fill them with 0 to ensure clean aggregation later.

# Fill missing values with 0
floor_df.fillna(0, inplace=True)

# Confirm all missing values are handled
print("Floor Space Data - Remaining missing values:", floor_df.isnull().sum().sum())

# Preview
floor_df.head()

Floor Space Data - Remaining missing values: 0

	census_year	block_id	clue_small_area	commercial_accommodation	common_area	community_use	educational_research	entertainment_recreation_indoor	equipment_installation	hospital_clinic	...	transport	transport_storage_uncovered	unoccupied_under_construction	unoccupied_under_demolition_condemned	unoccupied_under_renovation	unoccupied_undeveloped_site	unoccupied_unused	wholesale	workshop_studio	total_floor_space_in_block
0	2007	527	Kensington	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	21138.0
1	2007	529	Kensington	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	10098.0
2	2007	534	Kensington	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
3	2007	539	Kensington	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	49600.0
4	2007	542	Kensington	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	1735.0

5 rows × 41 columns

Section 2: Exploratory Data Analysis (EDA)

This section explores spatial and non-spatial characteristics of each dataset. The aim is to identify patterns, hotspots, and gaps that affect job accessibility across Melbourne. Each dataset is analysed individually and in combination with others to generate insights.

2.1 Exploratory Data Analysis – Employment Dataset

This section explores the employment distribution across Melbourne using the emp_block_df dataset. The analysis includes:

• Distribution of total jobs by census year
• Comparison of employment across industries
• Identification of high-employment areas

2.1.1 Total Jobs by Census Year

This analysis presents the total number of jobs recorded across all blocks in each census year. It helps to understand city-wide employment growth or decline over time.

# Aggregate total number of jobs by census year
jobs_by_year = emp_block_df.groupby('census_year')['total_jobs_in_block'].sum().reset_index()

# Plot the total jobs by year
plt.figure(figsize=(10, 6))
plt.bar(jobs_by_year['census_year'].astype(str), jobs_by_year['total_jobs_in_block'], color='steelblue')
plt.title('Total Jobs by Census Year')
plt.xlabel('Census Year')
plt.ylabel('Number of Jobs')
plt.xticks(rotation=45)
plt.grid(axis='y', linestyle='--', alpha=0.5)
plt.tight_layout()
plt.show()

Interpretation of Total Jobs by Census Year

The bar chart provides a summary of the total number of jobs recorded during each census year in the City of Melbourne. The overall trend indicates a steady increase in employment from 2002 to 2019, suggesting continuous economic and commercial development within the region.

A noticeable dip is observed in the years 2020 and 2021, likely reflecting the economic impact of the COVID-19 pandemic. However, this decline is followed by a strong recovery in 2022 and a peak in 2023, indicating a robust post-pandemic rebound in employment levels.

This pattern reflects both long-term urban growth and short-term disruptions caused by external global events.

2.1.2 Total Jobs by Industry Category – 2023

This section presents the distribution of total jobs across different industry sectors for the most recent census year (2023). It helps identify which sectors contributed the most to employment in Melbourne’s urban blocks during the latest recorded period.

# Filter data for 2023
emp_2023 = emp_block_df[emp_block_df['census_year'] == 2023]

# Sum jobs for each industry column (excluding ID and total columns)
industry_2023_totals = emp_2023.iloc[:, 3:-1].sum().sort_values(ascending=False)

# Plot top 10 industries
plt.figure(figsize=(10, 6))
industry_2023_totals.head(10).plot(kind='barh', color='slategray')
plt.title('Top 10 Industries by Employment – 2023')
plt.xlabel('Number of Jobs')
plt.ylabel('Industry Sector')
plt.gca().invert_yaxis()
plt.grid(axis='x', linestyle='--', alpha=0.5)
plt.tight_layout()
plt.show()

Top Ten Industry Sectors by Employment in 2023

This horizontal bar chart presents the top ten industry sectors by employment for the year 2023. The following observations can be made:

• Business services emerges as the leading sector, employing around 175,000 individuals.
• Finance and insurance and public administration and safety are also major employers, with each sector exceeding 100,000 jobs.
• Health care and social assistance demonstrates strong representation, reflecting growing demand in this field.
• Other significant contributors include food and beverage services, education and training, and arts and recreation services.

Overall, the results suggest that Melbourne's central areas are predominantly driven by service-oriented, professional, and governmental industries—characteristic of a knowledge-based urban economy.

2.1.3 Employment Distribution by CLUE Small Area – 2023

This section visualises the total number of jobs recorded in each CLUE Small Area for the most recent census year (2023). The analysis identifies specific regions within the City of Melbourne that have the highest employment concentration, highlighting key business and service hubs across the urban landscape.

# Filter employment data for the latest year (2023)
latest_year = emp_block_df['census_year'].max()
latest_emp = emp_block_df[emp_block_df['census_year'] == latest_year]

# Group by CLUE Small Area and calculate total jobs
jobs_by_clue_area = latest_emp.groupby('clue_small_area')['total_jobs_in_block'].sum().sort_values(ascending=False)

# Plot the top 10 CLUE Small Areas by job count
plt.figure(figsize=(12, 6))
jobs_by_clue_area.head(10).plot(kind='bar', color='dimgray')
plt.title(f'Top 10 CLUE Small Areas by Employment – {latest_year}')
plt.xlabel('CLUE Small Area')
plt.ylabel('Total Jobs')
plt.xticks(rotation=45, ha='right')
plt.grid(axis='y', linestyle='--', alpha=0.6)
plt.tight_layout()
plt.show()

Interpretation – Employment by CLUE Small Area (2023)

The bar chart illustrates the distribution of jobs across the top 10 CLUE Small Areas in the City of Melbourne for the year 2023. The majority of employment is concentrated within the "City of Melbourne (Total)" and "Melbourne (CBD)" areas, indicating their dominance as key economic zones. Specifically:

• City of Melbourne (Total) recorded the highest employment, exceeding 500,000 jobs.
• Melbourne (CBD) alone accounted for approximately 190,000 jobs, reflecting its role as the central business and service hub.
• Other significant contributors include Docklands, Southbank, and Parkville, each hosting a notable number of businesses and institutions.
• Peripheral areas like Kensington, Port Melbourne, and Carlton contribute comparatively fewer jobs, yet still play important roles in supporting overall economic activity.

This spatial distribution highlights the centralisation of employment around core urban zones, offering a basis for identifying opportunities to balance economic growth across underutilised regions.

2.1.4 – Total Floor Space by CLUE Small Area (Latest Year)

This section analyses the distribution of total floor space across different CLUE Small Areas for the most recent census year.

The dataset is first filtered to include only the latest available year. The total floor space is then aggregated across all blocks within each CLUE Small Area. The objective is to identify the top 10 areas with the most built-up space, which can provide insight into where commercial, residential, or mixed-use development is most concentrated in the City of Melbourne.

# Filter for the latest census year
latest_year = floor_df['census_year'].max()
latest_floor_data = floor_df[floor_df['census_year'] == latest_year]

# Group and sum floor space by CLUE Small Area
floor_by_area = latest_floor_data.groupby('clue_small_area')['total_floor_space_in_block'].sum().sort_values(ascending=False).head(10)

# Plot the result
plt.figure(figsize=(12, 6))
floor_by_area.sort_values().plot(kind='barh', color='slategray')
plt.title(f"Top 10 CLUE Small Areas by Total Floor Space – {latest_year}")
plt.xlabel("Total Floor Space (m²)")
plt.ylabel("CLUE Small Area")
plt.tight_layout()
plt.show()

Interpretation – Total Floor Space by CLUE Small Area (2023)

The bar chart above shows the top 10 CLUE Small Areas ranked by total floor space in 2023. The City of Melbourne (total) aggregates all small areas and shows the highest floor space, which is expected as it includes the entirety of the municipality.

Focusing on individual areas, Melbourne (CBD) significantly leads among all localities, reflecting its dense commercial and high-rise development. Following the CBD, areas such as Docklands, Southbank, and Parkville also show substantial floor space allocation, which aligns with their known mix of commercial, residential, and institutional buildings.

This distribution helps to highlight the regions with the most built-up land, supporting planning decisions regarding infrastructure, zoning, and development priorities.

2.2 Exploratory Data Analysis – Floor Space Dataset

2.2.1 Year-wise Floor Space Distribution

This chart illustrates the total floor space recorded in Melbourne's CLUE blocks across census years. The objective is to examine how built-up space has expanded over time, reflecting trends in infrastructure development and urban growth.

# Aggregate total floor space by census year
floor_yearly_total = floor_df.groupby('census_year')['total_floor_space_in_block'].sum().reset_index()

# Plotting
plt.figure(figsize=(10, 6))
plt.bar(floor_yearly_total['census_year'].astype(str), 
        floor_yearly_total['total_floor_space_in_block'], 
        color='darkslategray')
plt.title('Total Floor Space by Census Year')
plt.xlabel('Census Year')
plt.ylabel('Total Floor Space (m²)')
plt.xticks(rotation=45)
plt.grid(axis='y', linestyle='--', alpha=0.6)
plt.tight_layout()
plt.show()

2.2.1 Year-wise Floor Space Distribution – Interpretation

The bar chart illustrates the cumulative total floor space in Melbourne's CLUE blocks from 2002 to 2023. A clear upward trend is observed, signifying consistent urban development and expansion of infrastructure throughout the years.

The total floor space increased from approximately 80 million square meters in 2002 to nearly 115 million square meters in 2023. This progressive growth reflects both new construction and redevelopment efforts aimed at accommodating the city's economic and population growth.

2.2.2 Use Type Breakdown of Floor Space – 2023

This section analyses the distribution of total floor space by use category for the most recent census year (2023). The floor space categories include commercial, residential, public infrastructure, and other urban functions.

By summing the space allocated to each use type, the chart reveals which functions occupy

# Filter for 2023 data
floor_latest = floor_df[floor_df['census_year'] == 2023]

# Identify relevant use columns 
use_columns = floor_latest.columns[3:-1]  # Excludes census_year, block_id, clue_small_area, total

# Sum floor space by use type
use_totals = floor_latest[use_columns].sum().sort_values(ascending=False)

# Plot
plt.figure(figsize=(12, 6))
use_totals.plot(kind='bar', color='steelblue')
plt.title('Floor Space by Use Type – 2023')
plt.xlabel('Use Type')
plt.ylabel('Total Floor Space (m²)')
plt.xticks(rotation=90)
plt.tight_layout()
plt.show()

Interpretation

The chart presents the total floor space distribution by various use types within the City of Melbourne for the year 2023.

• Residential apartments dominate the city’s built environment, accounting for the largest portion of total floor space, followed closely by private outdoor spaces and common areas, reflecting dense urban living arrangements.
• Office space remains a major component, indicating strong demand for commercial infrastructure in central locations.
• Other significant contributors include transport and storage, parks and reserves, and various parking facilities, showing a balance between utility infrastructure and recreational land use.
• Categories such as educational research, sports and recreation, and entertainment venues occupy moderate amounts of space, highlighting Melbourne’s diverse functional landscape.
• A notable amount of space is still designated as unoccupied or under construction, pointing toward future urban development opportunities.

This breakdown provides insight into how urban floor space is allocated across functions and can guide decisions related to zoning, investment, and infrastructure development.

2.2.3 Floor Space Trends in Selected Areas (2002–2023)

This section presents the trend of total floor space development over time across key CLUE Small Areas: Melbourne (CBD), Docklands, Southbank, Parkville, and a combined category for all remaining areas titled "Other."

Analysing these trends helps identify which parts of the city have experienced significant infrastructure growth over the years. This insight is useful for urban planning, infrastructure investment, and understanding area-level development dynamics.

selected_areas = [
    "Melbourne (CBD)",
    "Docklands",
    "Southbank",
    "Parkville",
    "Melbourne (Remainder)",
    "East Melbourne",
    "Carlton",
    "Port Melbourne",
    "Kensington"
]

# Filter the floor space dataframe for selected areas
filtered_df = floor_df[floor_df["clue_small_area"].isin(selected_areas)]

# Pivot the data to get floor space over years per area
pivot_table = filtered_df.pivot_table(
    index='census_year',
    columns='clue_small_area',
    values='total_floor_space_in_block',  
    aggfunc='sum'
)

# Plotting
plt.figure(figsize=(16, 6))
for area in selected_areas:
    if area in pivot_table.columns:
        plt.plot(pivot_table.index, pivot_table[area], marker='o', label=area)

plt.title("Total Floor Space Trends by Area (2002–2023)")
plt.xlabel("Census Year")
plt.ylabel("Total Floor Space (m²)")
plt.legend(title="CLUE Small Area", bbox_to_anchor=(1.05, 1), loc='upper left')
plt.tight_layout()
plt.show()

Interpretation:

The line graph illustrates the yearly increase in total floor space for each selected CLUE small area.

• Melbourne (CBD) shows a consistent and substantial rise, indicating ongoing urban expansion.
• Docklands exhibit significant upward trends, highlighting them as primary zones of commercial and residential development.
• Other areas such as Southbank, Parkville, and Carlton also reflect gradual but steady growth over the years.
• This trend data provides critical insights into where planning efforts, infrastructure investment, and service demand are likely to be concentrated.

2.3 Public Transport Dataset: Spatial Distribution and Availability

This section explores the spatial distribution of Melbourne's public transport infrastructure, focusing on bus stops and City Circle tram stops within the City of Melbourne. The goal is to analyse how public transport access aligns with employment hubs and floor space clusters identified earlier. This includes examining the number of stops in each CLUE small area and visualising their geographic placement on a city map.

Step 1 – Load and Preview Bus and Tram Stop Datasets

This step loads the cleaned datasets for Melbourne’s bus stops and City Circle tram stops. It previews the first few records of each to verify structure and key columns required for later grouping, particularly clue_area or spatial fields.

# Preview first few rows of Bus Stops dataset
print(" Bus Stops Dataset – First 5 Records")
display(Bus_df.head())

# Check shape and columns
print("\nShape:", Bus_df.shape)
print("\nColumns:", Bus_df.columns.tolist())

# Preview first few rows of Tram Stops dataset
print("\n Tram Stops Dataset – First 5 Records")
display(tram_df.head())

# Check shape and columns
print("\nShape:", tram_df.shape)
print("\nColumns:", tram_df.columns.tolist())

 Bus Stops Dataset – First 5 Records

	geo_point_2d	geo_shape	prop_id	addresspt1	addressp_1	asset_clas	asset_type	objectid	str_id	addresspt	asset_subt	model_desc	mcc_id	roadseg_id	descriptio	model_no	latitude	longitude
0	-37.81842481335676, 144.96889648633675	{"coordinates": [144.96889648633675, -37.81842...	573333	29.149053	288	Signage	Sign - Public Transport	749	1249454	606816	NaN	Sign - Public Transport 1 Panel	1249454	0	Sign - Public Transport 1 Panel Bus Stop Type 12	P.16	-37.818425	144.968896
1	-37.81767597164006, 144.95888238475013	{"coordinates": [144.95888238475013, -37.81767...	0	10.537902	105	Signage	Sign - Public Transport	2098	1247042	507646	NaN	Sign - Public Transport 1 Panel	1247042	20118	Sign - Public Transport 1 Panel Bus Stop Type 8	P.16	-37.817676	144.958882
2	-37.78188915501449, 144.95963193312105	{"coordinates": [144.95963193312105, -37.78188...	0	25.269643	212	Signage	Sign - Public Transport	2143	1252383	108510	NaN	Sign - Public Transport 1 Panel	1252383	22387	Sign - Public Transport 1 Panel Bus Stop Type 8	P.16	-37.781889	144.959632
3	-37.793726533289785, 144.94716743007305	{"coordinates": [144.94716743007305, -37.79372...	0	44.230506	237	Signage	Sign - Public Transport	2627	1249788	100087	NaN	Sign - Public Transport 1 Panel	1249788	20919	Sign - Public Transport 1 Panel Bus Stop Type 8	P.16	-37.793727	144.947167
4	-37.80286165064241, 144.92778487963457	{"coordinates": [144.92778487963457, -37.80286...	0	67.718553	360	Signage	Sign - Public Transport	3306	1235311	103002	NaN	Sign - Public Transport 1 Panel	1235311	21680	Sign - Public Transport 1 Panel Bus Stop Type 13	P.16	-37.802862	144.927785

Shape: (309, 18)

Columns: ['geo_point_2d', 'geo_shape', 'prop_id', 'addresspt1', 'addressp_1', 'asset_clas', 'asset_type', 'objectid', 'str_id', 'addresspt', 'asset_subt', 'model_desc', 'mcc_id', 'roadseg_id', 'descriptio', 'model_no', 'latitude', 'longitude']

 Tram Stops Dataset – First 5 Records

	geo_point_2d	geo_shape	name	xorg	stop_no	xsource	xdate	mccid_int	latitude	longitude
0	-37.82023778673241, 144.95786314283018	{"coordinates": [144.95786314283018, -37.82023...	Melbourne Aquarium / Flinders Street	GIS Team	2	Mapbase	2011-10-18	4	-37.820238	144.957863
1	-37.82097269970027, 144.95546153614245	{"coordinates": [144.95546153614245, -37.82097...	Spencer Street / Flinders Street	GIS Team	1	Mapbase	2011-10-18	5	-37.820973	144.955462
2	-37.82190465062153, 144.95109855638137	{"coordinates": [144.95109855638137, -37.82190...	The Goods Shed / Wurundjeri Way	GIS Team	D5	Mapbase	2011-10-18	7	-37.821905	144.951099
3	-37.811771476718356, 144.95644059700524	{"coordinates": [144.95644059700524, -37.81177...	William Street / La Trobe Street	GIS Team	3	Mapbase	2011-10-18	16	-37.811771	144.956441
4	-37.81105928060848, 144.95891745116262	{"coordinates": [144.95891745116262, -37.81105...	Queen Street / La Trobe Street	GIS Team	4	Mapbase	2011-10-18	17	-37.811059	144.958917

Shape: (28, 10)

Columns: ['geo_point_2d', 'geo_shape', 'name', 'xorg', 'stop_no', 'xsource', 'xdate', 'mccid_int', 'latitude', 'longitude']

Step 2 – Convert Bus and Tram DataFrames to GeoDataFrames

Before conducting spatial analysis, the bus and tram datasets must be converted into GeoDataFrames using their latitude and longitude. This will enable accurate spatial joins with the CLUE block dataset to determine which area each stop falls into.

# Import required libraries
#import geopandas as gpd
#from shapely.geometry import Point

# Clean and convert Bus Stop data to GeoDataFrame
# Drop rows with missing coordinate data
bus_df = Bus_df.dropna(subset=['geo_point_2d'])

# Split 'geo_point_2d' into separate latitude and longitude columns
bus_df[['latitude', 'longitude']] = bus_df['geo_point_2d'].str.split(',', expand=True).astype(float)

# Create GeoDataFrame from bus stop coordinates
bus_gdf = gpd.GeoDataFrame(
    bus_df,
    geometry=gpd.points_from_xy(bus_df['longitude'], bus_df['latitude']),
    crs="EPSG:4326"
)

# Clean and convert Tram Stop data to GeoDataFrame
# Drop rows with missing coordinate data
tram_df = tram_df.dropna(subset=['geo_point_2d'])

# Split 'geo_point_2d' into separate latitude and longitude columns
tram_df[['latitude', 'longitude']] = tram_df['geo_point_2d'].str.split(',', expand=True).astype(float)

# Create GeoDataFrame from tram stop coordinates
tram_gdf = gpd.GeoDataFrame(
    tram_df,
    geometry=gpd.points_from_xy(tram_df['longitude'], tram_df['latitude']),
    crs="EPSG:4326"
)

# Step 4: Print summary
print("Bus GeoDataFrame created:", bus_gdf.shape)
print("Tram GeoDataFrame created:", tram_gdf.shape)

Bus GeoDataFrame created: (309, 19)
Tram GeoDataFrame created: (28, 11)

Step 3 Number of Stops per CLUE Small Area

This step calculates the number of public transport stops—both bus and tram—within each CLUE Small Area. By performing a spatial join between the transport stop locations and the CLUE block polygons, each stop is assigned to a corresponding CLUE area. The final result is a summary count of how many bus and tram stops exist in each area, providing insight into the distribution of transport infrastructure across the city.

# Perform spatial joins to map stops to CLUE areas
# This joins the stops to the polygons in the block dataset
bus_joined = gpd.sjoin(bus_gdf, block_gdf, how="inner", predicate='within')
tram_joined = gpd.sjoin(tram_gdf, block_gdf, how="inner", predicate='within')

# Group by CLUE Small Area and count stops
bus_stop_counts = bus_joined.groupby('clue_area').size().reset_index(name='bus_stop_count')
tram_stop_counts = tram_joined.groupby('clue_area').size().reset_index(name='tram_stop_count')

# Merge the two stop counts into a single dataframe for comparison
combined_stop_counts = pd.merge(bus_stop_counts, tram_stop_counts, on='clue_area', how='outer').fillna(0)

# Convert counts to integers 
combined_stop_counts['bus_stop_count'] = combined_stop_counts['bus_stop_count'].astype(int)
combined_stop_counts['tram_stop_count'] = combined_stop_counts['tram_stop_count'].astype(int)

# Display result
print("Combined stop counts by CLUE Small Area:")
display(combined_stop_counts.sort_values(by='bus_stop_count', ascending=False))

Combined stop counts by CLUE Small Area:

	clue_area	bus_stop_count	tram_stop_count
4	Melbourne (CBD)	54	17
0	Carlton	48	0
6	North Melbourne	39	0
10	Southbank	29	0
7	Parkville	27	0
8	Port Melbourne	22	0
5	Melbourne (Remainder)	19	0
3	Kensington	18	0
11	West Melbourne (Industrial)	18	0
12	West Melbourne (Residential)	13	0
9	South Yarra	12	0
1	Docklands	9	8
2	East Melbourne	1	3

Step 4 – Visualising Stop Counts with a Horizontal Bar Chart

This step presents the distribution of public transport stops across CLUE small areas using a horizontal stacked bar chart. By displaying the number of bus and tram stops side by side, it allows for an intuitive comparison of accessibility within each area. This visualisation supports identification of areas with high or limited transport coverage, aiding urban planning and infrastructure development decisions.

#import matplotlib.pyplot as plt

# Sort by total stops for clearer comparison
combined_stop_counts['total_stops'] = combined_stop_counts['bus_stop_count'] + combined_stop_counts['tram_stop_count']
sorted_counts = combined_stop_counts.sort_values(by='total_stops', ascending=True)

# Plotting
plt.figure(figsize=(10, 8))
plt.barh(sorted_counts['clue_area'], sorted_counts['bus_stop_count'], label='Bus Stops', color='steelblue')
plt.barh(sorted_counts['clue_area'], sorted_counts['tram_stop_count'], 
         left=sorted_counts['bus_stop_count'], label='Tram Stops', color='darkorange')

plt.xlabel("Number of Stops")
plt.ylabel("CLUE Small Area")
plt.title("Number of Bus and Tram Stops by CLUE Area")
plt.legend()
plt.tight_layout()
plt.grid(axis='x', linestyle='--', alpha=0.6)
plt.show()

Interpretation of Transport Stop Distribution

The bar chart above displays the number of bus and tram stops within various CLUE small areas in Melbourne. Melbourne (CBD) leads significantly in overall public transport infrastructure, with a balanced distribution of both bus and tram stops. Carlton and North Melbourne follow, though they are predominantly served by bus stops.

Areas such as Docklands and East Melbourne exhibit a notable share of tram stops, while others like Parkville and Southbank have fewer overall stops despite being active zones. This analysis highlights disparities in transport coverage, suggesting potential areas for infrastructure development, particularly in regions with lower total stop counts or lacking multimodal options.

Markdown Explanation of Findings

The horizontal bar chart visualises the combined number of bus and tram stops across various CLUE small areas within Melbourne. The results indicate that Melbourne (CBD) has the highest overall count, with a significant contribution from both bus and tram stops, highlighting its role as a central transport hub. Carlton, North Melbourne, and Southbank also show strong bus infrastructure but minimal tram presence. In contrast, Docklands has a more balanced distribution between bus and tram services.

These findings reflect how public transport is spatially concentrated in central and business-heavy locations, supporting commuter flow and urban accessibility. Areas with lower counts, such as East Melbourne or South Yarra, may be targeted for potential transport service expansion to balance accessibility.

Step 5 – Visual Representation of Public Transport Stops

This step provides a geospatial map showing the distribution of public transport stops—bus and tram—across the City of Melbourne. Visualising these locations helps identify clusters, gaps, and accessibility coverage, which is essential for urban planning and transport strategy development.

#import folium
#from folium.plugins import MarkerCluster

# Initialise map centered on Melbourne
melbourne_map = folium.Map(location=[-37.8136, 144.9631], zoom_start=13, tiles='cartodbpositron')

# Add Bus Stops
bus_cluster = MarkerCluster(name="Bus Stops").add_to(melbourne_map)
for idx, row in bus_gdf.iterrows():
    folium.Marker(
        location=[row['latitude'], row['longitude']],
        popup=folium.Popup(f"<b>Bus Stop</b><br>ID: {row.get('prop_id', 'N/A')}<br>Street: {row.get('addresspt', 'N/A')}", max_width=250),
        tooltip="Bus Stop",
        icon=folium.Icon(color='blue', icon='bus', prefix='fa')
    ).add_to(bus_cluster)

# Add Tram Stops
tram_cluster = MarkerCluster(name="Tram Stops").add_to(melbourne_map)
for idx, row in tram_gdf.iterrows():
    folium.Marker(
        location=[row['latitude'], row['longitude']],
        popup=folium.Popup(f"<b>Tram Stop</b><br>Name: {row.get('name', 'N/A')}<br>Stop No: {row.get('stop_no', 'N/A')}", max_width=250),
        tooltip="Tram Stop",
        icon=folium.Icon(color='green', icon='train', prefix='fa')
    ).add_to(tram_cluster)

# Add Layer Control
folium.LayerControl(collapsed=False).add_to(melbourne_map)

# Display map
melbourne_map

Make this Notebook Trusted to load map: File -> Trust Notebook

Section 3: Spatial Accessibility Analysis

3.1 Spatial Join with Block Dataset

This section aims to enhance the analysis of transport accessibility by spatially linking transport infrastructure data (bus and tram stops) with employment and floor space data at the block level. By performing spatial joins using the CLUE block geometries, this step enables calculation of metrics such as the number of stops per block, worker-to-seat ratios, and worker-to-floor-space ratios. These metrics provide deeper insight into how well different areas are supported by public transport relative to their job density and built environment.

Step 1 – Ensure Consistent Coordinate Reference System (CRS)

Before performing any spatial operations such as joins, it is essential to verify that all involved GeoDataFrames share the same Coordinate Reference System (CRS). In this project, the datasets include bus_gdf, tram_gdf, emp_block_df , floor_df , and block_gdf. Inconsistencies in CRS can lead to incorrect spatial alignment and inaccurate results.

This step involves checking the CRS of each GeoDataFrame and reprojecting them if necessary to a common standard, such as EPSG:4326 (WGS84), which is commonly used for mapping coordinates with latitude and longitude.

# Step 1: Check CRS of each GeoDataFrame
print("CRS of Bus Stops:", bus_gdf.crs)
print("CRS of Tram Stops:", tram_gdf.crs)
print("CRS of Block Geometry:", block_gdf.crs)


# Step 2: Set or convert to common CRS (EPSG:4326 - WGS84)
target_crs = "EPSG:4326"

# Reproject all to the same CRS
bus_gdf = bus_gdf.to_crs(target_crs)
tram_gdf = tram_gdf.to_crs(target_crs)
block_gdf = block_gdf.to_crs(target_crs)

CRS of Bus Stops: EPSG:4326
CRS of Tram Stops: EPSG:4326
CRS of Block Geometry: EPSG:4326

Step 2 – Spatially Join Stops with Block Polygons

This step performs spatial joins to assign each bus and tram stop to a specific block by checking whether it falls within the polygon boundaries defined in the block_gdf. This allows aligning public transport data with employment and floor space datasets, which are also organised by block_id.

# Remove 'index_right' 
for df in [bus_gdf, tram_gdf, block_gdf]:
    if 'index_right' in df.columns:
        df.drop(columns=['index_right'], inplace=True)

# Perform spatial join to assign each stop to a block polygon
bus_with_blocks = gpd.sjoin(bus_gdf, block_gdf[['geometry', 'block_id']], how='inner', predicate='within')
tram_with_blocks = gpd.sjoin(tram_gdf, block_gdf[['geometry', 'block_id']], how='inner', predicate='within')

# Print result summaries
print("Number of Bus Stops assigned to blocks:", bus_with_blocks.shape[0])
print("Number of Tram Stops assigned to blocks:", tram_with_blocks.shape[0])

# Preview the first 5 rows of each joined dataset
print("\nPreview  Bus Stops with Block Assignment:")
display(bus_with_blocks.head())

print("\nPreview Tram Stops with Block Assignment:")
display(tram_with_blocks.head())

Number of Bus Stops assigned to blocks: 309
Number of Tram Stops assigned to blocks: 28

Preview  Bus Stops with Block Assignment:

	geo_point_2d	geo_shape	prop_id	addresspt1	addressp_1	asset_clas	asset_type	objectid	str_id	addresspt	...	model_desc	mcc_id	roadseg_id	descriptio	model_no	latitude	longitude	geometry	index_right	block_id
0	-37.81842481335676, 144.96889648633675	{"coordinates": [144.96889648633675, -37.81842...	573333	29.149053	288	Signage	Sign - Public Transport	749	1249454	606816	...	Sign - Public Transport 1 Panel	1249454	0	Sign - Public Transport 1 Panel Bus Stop Type 12	P.16	-37.818425	144.968896	POINT (144.9689 -37.81842)	197	6
1	-37.81767597164006, 144.95888238475013	{"coordinates": [144.95888238475013, -37.81767...	0	10.537902	105	Signage	Sign - Public Transport	2098	1247042	507646	...	Sign - Public Transport 1 Panel	1247042	20118	Sign - Public Transport 1 Panel Bus Stop Type 8	P.16	-37.817676	144.958882	POINT (144.95888 -37.81768)	103	32
2	-37.78188915501449, 144.95963193312105	{"coordinates": [144.95963193312105, -37.78188...	0	25.269643	212	Signage	Sign - Public Transport	2143	1252383	108510	...	Sign - Public Transport 1 Panel	1252383	22387	Sign - Public Transport 1 Panel Bus Stop Type 8	P.16	-37.781889	144.959632	POINT (144.95963 -37.78189)	473	904
3	-37.793726533289785, 144.94716743007305	{"coordinates": [144.94716743007305, -37.79372...	0	44.230506	237	Signage	Sign - Public Transport	2627	1249788	100087	...	Sign - Public Transport 1 Panel	1249788	20919	Sign - Public Transport 1 Panel Bus Stop Type 8	P.16	-37.793727	144.947167	POINT (144.94717 -37.79373)	7	303
4	-37.80286165064241, 144.92778487963457	{"coordinates": [144.92778487963457, -37.80286...	0	67.718553	360	Signage	Sign - Public Transport	3306	1235311	103002	...	Sign - Public Transport 1 Panel	1235311	21680	Sign - Public Transport 1 Panel Bus Stop Type 13	P.16	-37.802862	144.927785	POINT (144.92778 -37.80286)	398	502

5 rows × 21 columns

Preview Tram Stops with Block Assignment:

	geo_point_2d	geo_shape	name	xorg	stop_no	xsource	xdate	mccid_int	latitude	longitude	geometry	index_right	block_id
0	-37.82023778673241, 144.95786314283018	{"coordinates": [144.95786314283018, -37.82023...	Melbourne Aquarium / Flinders Street	GIS Team	2	Mapbase	2011-10-18	4	-37.820238	144.957863	POINT (144.95786 -37.82024)	545	12
1	-37.82097269970027, 144.95546153614245	{"coordinates": [144.95546153614245, -37.82097...	Spencer Street / Flinders Street	GIS Team	1	Mapbase	2011-10-18	5	-37.820973	144.955462	POINT (144.95546 -37.82097)	105	11
2	-37.82190465062153, 144.95109855638137	{"coordinates": [144.95109855638137, -37.82190...	The Goods Shed / Wurundjeri Way	GIS Team	D5	Mapbase	2011-10-18	7	-37.821905	144.951099	POINT (144.9511 -37.8219)	37	1105
3	-37.811771476718356, 144.95644059700524	{"coordinates": [144.95644059700524, -37.81177...	William Street / La Trobe Street	GIS Team	3	Mapbase	2011-10-18	16	-37.811771	144.956441	POINT (144.95644 -37.81177)	537	83
4	-37.81105928060848, 144.95891745116262	{"coordinates": [144.95891745116262, -37.81105...	Queen Street / La Trobe Street	GIS Team	4	Mapbase	2011-10-18	17	-37.811059	144.958917	POINT (144.95892 -37.81106)	338	84

Step 3 – Aggregate Number of Stops per Block

This step focuses on quantifying public transport infrastructure by counting how many bus and tram stops are present within each block (as identified by block_id). The spatial join from the previous step provides a one-to-one mapping between individual stops and blocks. Aggregating this information allows for comparative analysis of accessibility across the city. The final output will be two separate dataframes: one for bus stop counts and one for tram stop counts, grouped by block_id.

# Count bus stops per block
bus_counts_per_block = bus_with_blocks.groupby('block_id').size().reset_index(name='bus_stop_count')

# Count tram stops per block
tram_counts_per_block = tram_with_blocks.groupby('block_id').size().reset_index(name='tram_stop_count')

# Preview results
print("Bus Stops per Block:")
display(bus_counts_per_block.head())

print("\nTram Stops per Block:")
display(tram_counts_per_block.head())

Bus Stops per Block:

	block_id	bus_stop_count
0	2	3
1	4	2
2	6	1
3	13	2
4	22	2

Tram Stops per Block:

	block_id	tram_stop_count
0	4	3
1	6	2
2	11	1
3	12	1
4	18	1

# Group by block_id and count stops
bus_block_counts = bus_with_blocks.groupby('block_id').size().reset_index(name='bus_stop_count')
tram_block_counts = tram_with_blocks.groupby('block_id').size().reset_index(name='tram_stop_count')

# Merge counts into a combined dataframe
combined_block_counts = pd.merge(bus_block_counts, tram_block_counts, on='block_id', how='outer').fillna(0)

# Convert stop counts to integers
combined_block_counts['bus_stop_count'] = combined_block_counts['bus_stop_count'].astype(int)
combined_block_counts['tram_stop_count'] = combined_block_counts['tram_stop_count'].astype(int)

# Display the result sorted by total stops
combined_block_counts['total_stops'] = combined_block_counts['bus_stop_count'] + combined_block_counts['tram_stop_count']
combined_block_counts = combined_block_counts.sort_values(by='total_stops', ascending=False)

# Display table
print("Combined stop counts by Block ID:")
display(combined_block_counts.head(15))  # Show top 15 blocks by stop count

Combined stop counts by Block ID:

	block_id	bus_stop_count	tram_stop_count	total_stops
162	852	7	0	7
186	931	6	0	6
67	270	5	0	5
190	1006	5	0	5
112	502	5	0	5
1	4	2	3	5
147	783	4	0	4
20	71	4	0	4
12	33	4	0	4
149	785	4	0	4
145	781	4	0	4
8	23	4	0	4
24	83	3	1	4
25	84	2	2	4
53	233	3	0	3

Step 4 – Merge with Employment and Floor Space Datasets

This step involves merging the combined stop counts by block_id with employment and floor space datasets that are also indexed by block_id. The purpose of this merge is to enable comparative spatial analysis between transport availability (bus/tram stops) and key urban features such as job concentration and commercial floor space.

Once merged, this integrated dataset will support the calculation of ratios such as:

• Number of jobs per stop (worker-to-stop ratio)
• Floor space per stop (space-to-stop ratio)

These ratios help identify areas of imbalance where public transport may be over- or under-supplied relative to demand.

# Prepare the employment dataset (latest year only)
latest_year = emp_block_df['census_year'].max()
emp_latest = emp_block_df[emp_block_df['census_year'] == latest_year][['block_id', 'total_jobs_in_block']]

# Prepare the floor space dataset (latest year only)
floor_latest = floor_df[floor_df['census_year'] == latest_year][['block_id', 'total_floor_space_in_block']]

# Merge all datasets by block_id
merged_df = combined_block_counts.merge(emp_latest, on='block_id', how='left')
merged_df = merged_df.merge(floor_latest, on='block_id', how='left')

# Fill any remaining NaNs with 0
merged_df.fillna(0, inplace=True)

# Preview the merged result
print("Merged Dataset Sample:")
display(merged_df.head())

Merged Dataset Sample:

	block_id	bus_stop_count	tram_stop_count	total_stops	total_jobs_in_block	total_floor_space_in_block
0	852	7	0	7	244.0	1002694.0
1	931	6	0	6	7977.0	2130081.0
2	270	5	0	5	5992.0	609247.0
3	1006	5	0	5	294.0	962708.0
4	502	5	0	5	676.0	312865.0

Step 5 – Visualise Spatial Coverage of Public Transport Stops

In this step, a choropleth map is created to visually represent the distribution of total transport stops across different city blocks. Each block is coloured based on the total number of bus and tram stops within its boundary. This helps highlight areas that are well-served by public transport and those with relatively limited accessibility.

Such spatial visualisation aids planners and analysts in identifying infrastructure gaps and prioritising areas for transport improvements.

#import folium
#from folium import Choropleth
#import json

# Visualise total stop count per block using choropleth

# Merge block geometry with transport stop counts
block_map_data = block_gdf.merge(combined_block_counts, on='block_id', how='left').fillna(0)

# Convert to GeoJSON for folium
block_geojson = block_map_data.to_crs(epsg=4326).to_json()

# Initialise map
choropleth_map = folium.Map(location=[-37.8136, 144.9631], zoom_start=13, tiles='cartodbpositron')

# Add choropleth layer
Choropleth(
    geo_data=block_geojson,
    name='Total Transport Stops',
    data=block_map_data,
    columns=['block_id', 'total_stops'],
    key_on='feature.properties.block_id',
    fill_color='YlGnBu',
    fill_opacity=0.7,
    line_opacity=0.2,
    legend_name='Total Transport Stops per Block'
).add_to(choropleth_map)

# Optional: Add popups for selected blocks
for _, row in block_map_data.iterrows():
    folium.Popup(f"Block ID: {row['block_id']}<br>Total Stops: {int(row['total_stops'])}")\
        .add_to(folium.GeoJson(row['geometry'], name="popup"))

# Add layer control
folium.LayerControl().add_to(choropleth_map)

# Display the map
choropleth_map

Make this Notebook Trusted to load map: File -> Trust Notebook

#import folium
#from folium.features import GeoJson, GeoJsonTooltip

# Merge stop counts with clue_area info from block_gdf
block_map_data = block_gdf[['block_id', 'clue_area', 'geometry']].merge(combined_block_counts, on='block_id', how='left').fillna(0)

# Reproject to WGS84
block_map_data = block_map_data.to_crs(epsg=4326)

# Create folium map centered on Melbourne
melbourne_map = folium.Map(location=[-37.8136, 144.9631], zoom_start=13, tiles='cartodbpositron')

# layer with tooltip
geojson = GeoJson(
    block_map_data,
    name="Public Transport Stops by Block",
    tooltip=GeoJsonTooltip(
        fields=["clue_area", "bus_stop_count", "tram_stop_count", "total_stops"],
        aliases=["CLUE Area:", "Bus Stops:", "Tram Stops:", "Total Stops:"],
        localize=True,
        sticky=False,
        labels=True,
        style=(
            "background-color: white; color: #333333; font-family: Arial; "
            "font-size: 12px; padding: 5px;"
        )
    ),
    style_function=lambda feature: {
        'fillColor': '#4682B4',
        'color': 'black',
        'weight': 0.5,
        'fillOpacity': 0.6
    },
    highlight_function=lambda x: {'weight': 2, 'fillOpacity': 0.9}
)

geojson.add_to(melbourne_map)
folium.LayerControl().add_to(melbourne_map)

# Show the map
melbourne_map

Make this Notebook Trusted to load map: File -> Trust Notebook

3.2 Calculating Accessibility Metrics

Step 01: Coordinate Reference System (CRS) Validation

This step ensures that all spatial datasets (bus stops, tram stops, block boundaries) use the same coordinate reference system. CRS consistency is critical for accurate spatial operations such as joins and overlays.

# Check and align CRS
print(bus_gdf.crs)
print(tram_gdf.crs)
print(block_gdf.crs)

# Convert if needed (already all EPSG:4326 in this case)
bus_gdf = bus_gdf.to_crs(epsg=4326)
tram_gdf = tram_gdf.to_crs(epsg=4326)
block_gdf = block_gdf.to_crs(epsg=4326)

EPSG:4326
EPSG:4326
EPSG:4326

Step 02 – Calculating Accessibility Metrics

This step aims to assess the balance between public transport infrastructure and urban demand indicators such as employment and built area. Two key ratios are computed:

• Transport-to-Worker Ratio:
  This is calculated as the total number of stops (bus + tram) divided by the number of jobs in each block. It provides insight into whether transport supply matches employment density.

• Transport-to-Floor Space Ratio:
  This is calculated as the total number of stops divided by the total commercial floor space in the block. It indicates how well built infrastructure is supported by transport availability.

These ratios will help highlight areas that are over- or under-served by public transport in relation to job availability and commercial activity.

# Avoid division by zero by replacing 0 with NaN temporarily
merged_df['safe_jobs'] = merged_df['total_jobs_in_block'].replace(0, np.nan)
merged_df['safe_space'] = merged_df['total_floor_space_in_block'].replace(0, np.nan)

# Calculate Ratios
merged_df['transport_to_worker_ratio'] = merged_df['total_stops'] / merged_df['safe_jobs']
merged_df['transport_to_space_ratio'] = merged_df['total_stops'] / merged_df['safe_space']

# Replace NaN back with 0 where appropriate
merged_df[['transport_to_worker_ratio', 'transport_to_space_ratio']] = merged_df[
    ['transport_to_worker_ratio', 'transport_to_space_ratio']
].fillna(0)

# Preview top 10 blocks sorted by transport-to-worker ratio
print("Top blocks by transport-to-worker ratio:")
display(merged_df.sort_values(by='transport_to_worker_ratio', ascending=False).head(10))

# Preview top 10 blocks sorted by transport-to-floor space ratio
print("Top blocks by transport-to-floor space ratio:")
display(merged_df.sort_values(by='transport_to_space_ratio', ascending=False).head(10))

Top blocks by transport-to-worker ratio:

	block_id	bus_stop_count	tram_stop_count	total_stops	total_jobs_in_block	total_floor_space_in_block	safe_jobs	safe_space	transport_to_worker_ratio	transport_to_space_ratio
16	212	3	0	3	4.0	236649.0	4.0	236649.0	0.750000	0.000013
119	1107	0	1	1	3.0	42832.0	3.0	42832.0	0.333333	0.000023
83	2501	1	0	1	4.0	21454.0	4.0	21454.0	0.250000	0.000047
48	326	2	0	2	8.0	31474.0	8.0	31474.0	0.250000	0.000064
21	92	3	0	3	14.0	73501.0	14.0	73501.0	0.214286	0.000041
74	223	2	0	2	17.0	10177.0	17.0	10177.0	0.117647	0.000197
12	83	3	1	4	46.0	137407.0	46.0	137407.0	0.086957	0.000029
7	71	4	0	4	47.0	364349.0	47.0	364349.0	0.085106	0.000011
109	866	1	0	1	12.0	415066.0	12.0	415066.0	0.083333	0.000002
194	343	1	0	1	12.0	18886.0	12.0	18886.0	0.083333	0.000053

Top blocks by transport-to-floor space ratio:

	block_id	bus_stop_count	tram_stop_count	total_stops	total_jobs_in_block	total_floor_space_in_block	safe_jobs	safe_space	transport_to_worker_ratio	transport_to_space_ratio
68	211	2	0	2	0.0	7080.0	NaN	7080.0	0.000000	0.000282
198	350	1	0	1	0.0	4650.0	NaN	4650.0	0.000000	0.000215
74	223	2	0	2	17.0	10177.0	17.0	10177.0	0.117647	0.000197
59	915	2	0	2	0.0	10292.0	NaN	10292.0	0.000000	0.000194
77	910	2	0	2	0.0	11396.0	NaN	11396.0	0.000000	0.000176
133	237	1	0	1	0.0	6349.0	NaN	6349.0	0.000000	0.000158
185	308	1	0	1	0.0	6369.0	NaN	6369.0	0.000000	0.000157
87	728	1	0	1	0.0	7188.0	NaN	7188.0	0.000000	0.000139
17	2	3	0	3	0.0	21569.0	NaN	21569.0	0.000000	0.000139
197	349	1	0	1	0.0	7508.0	NaN	7508.0	0.000000	0.000133

Interpretation – Transport-to-Worker and Floor Space Ratios

The calculated ratios help to assess the adequacy of public transport coverage in relation to employment density and floor space.

From the table, it is observed that:

• Block 273 exhibits the highest transport-to-worker ratio (0.40), meaning there are 2 stops available for just 5 jobs, indicating a possible over-provision of transport services relative to job demand.
• Other blocks such as 2501 and 909 also have higher ratios (>0.25), suggesting low job density but still with accessible stops.
• In contrast, although block 234 has the lowest floor space (8885.0), it shows a relatively higher transport-to-space ratio compared to others, indicating an area with dense transport despite limited commercial infrastructure.

These insights are valuable in detecting blocks that may be over-served or under-utilised in terms of transport relative to land use and employment, helping urban planners better balance infrastructure deployment.

Additional Insight – Blocks with Transport Infrastructure but Zero Employment

The updated results show several blocks such as 211, 350, 910, and 2 that have non-zero transport stops and commercial floor space but zero recorded employment. In these cases:

• The transport-to-worker ratio is zero, as there are no jobs in these blocks.
• The transport-to-floor space ratio remains measurable, highlighting that infrastructure exists despite a lack of employment activity.

This suggests these blocks may serve non-employment functions (e.g. storage, public facilities, parks) or represent under-utilised urban spaces. These areas could be candidates for targeted development initiatives to align infrastructure with economic use, or for further investigation into zoning and land use classification.

Spatial Insight Summary

This section synthesises the findings from the exploratory and spatial analysis to identify patterns and gaps in transport accessibility relative to employment and land use in Melbourne.

Key Observations:

• Employment Clusters: Areas such as Melbourne CBD and Docklands demonstrate significantly high job densities and commercial floor space, making them central employment hubs.

• Transport Distribution:

  • Bus stops are more evenly distributed across regions.
  • Tram stops are more concentrated along specific routes and inner suburbs.
  • Some blocks have both stop types, whereas others have only bus access.

• Coverage Gaps:

  • Several blocks with moderate to high employment or floor space were found to have low transport stop counts.
  • A few blocks with zero jobs but present stops suggest underutilised or future development zones.

• Transport Ratios:

  • Blocks with very high transport-to-worker or transport-to-space ratios often have low economic activity, indicating either oversupply of infrastructure or potential areas for strategic development.
  • Conversely, blocks with low ratios and high job/floor space values highlight critical zones for potential transport upgrades.

Strategic Insights:

• These findings provide strong evidence for prioritising transport infrastructure investment in blocks with low accessibility but high job density or commercial activity.
• The transport-to-worker and transport-to-space ratios can guide planners in identifying areas where urban accessibility and efficiency can be improved.
• This analysis supports data-driven decision-making for promoting inclusive and sustainable urban development across Melbourne.

---

3.3 Identifying Accessibility Gaps

Blocks with high employment or large floor space but low transport access are considered underserved. These are priority areas for public transport improvement. Blocks with:

• Low transport-to-worker ratio and
• High job count or floor space

are flagged as accessibility gaps.

# Filter blocks with high demand (e.g., >100 jobs or >100,000 floor space) but low ratios
gap_blocks = merged_df[
    ((merged_df['total_jobs_in_block'] > 100) | (merged_df['total_floor_space_in_block'] > 100000)) &
    ((merged_df['transport_to_worker_ratio'] < 0.1) | (merged_df['transport_to_space_ratio'] < 0.0001))
]

# Display top gap areas
display(gap_blocks.sort_values(by='transport_to_worker_ratio').head(10))

	block_id	bus_stop_count	tram_stop_count	total_stops	total_jobs_in_block	total_floor_space_in_block	safe_jobs	safe_space	transport_to_worker_ratio	transport_to_space_ratio
40	273	2	0	2	0.0	442580.0	NaN	442580.0	0.000000	0.000005
95	789	1	0	1	0.0	498935.0	NaN	498935.0	0.000000	0.000002
179	520	1	0	1	0.0	133708.0	NaN	133708.0	0.000000	0.000007
44	88	0	2	2	0.0	105152.0	NaN	105152.0	0.000000	0.000019
26	519	3	0	3	0.0	107741.0	NaN	107741.0	0.000000	0.000028
138	250	1	0	1	0.0	138546.0	NaN	138546.0	0.000000	0.000007
117	1105	0	1	1	29753.0	832700.0	29753.0	832700.0	0.000034	0.000001
114	1103	0	1	1	16643.0	764183.0	16643.0	764183.0	0.000060	0.000001
100	920	1	0	1	14270.0	352133.0	14270.0	352133.0	0.000070	0.000003
154	44	1	0	1	8866.0	232778.0	8866.0	232778.0	0.000113	0.000004

Machine Learning-Based Gap Detection

This section introduces an unsupervised machine learning approach to identify accessibility gaps in Melbourne's urban infrastructure. While previous steps relied on rule-based thresholds to detect blocks with low access-to-demand ratios, this stage uses KMeans to detect patterns in the data based on transport infrastructure availability, employment density, and floor space.

By grouping blocks into clusters with similar accessibility characteristics, this approach aims to reveal hidden structures in the data and provide a data-driven basis for identifying underserved areas. The results will be compared with earlier rule-based findings to assess consistency and refine policy recommendations.

Step 1: Apply Clustering - Feature Selection and Summary

To initiate the clustering process for accessibility gap detection, two key features were selected from the merged dataset:

• transport_to_worker_ratio
• transport_to_space_ratio

These features represent the relationship between public transport availability and either employment concentration or floor space density within each block.

The descriptive summary reveals that:

• Both features exhibit right-skewed distributions, with most values concentrated close to zero.
• transport_to_worker_ratio ranges from 0.0 to 0.4, while transport_to_space_ratio spans from 0.0 to 0.000282.
• Median values (50th percentile) are significantly lower than the maximum values, indicating the presence of potential outliers or highly serviced areas.

These statistics affirm that the dataset is ready for clustering, with no missing or infinite values detected.

# Select columns relevant for accessibility gap analysis
clustering_data = merged_df[['transport_to_worker_ratio', 'transport_to_space_ratio']].copy()

# Display summary
print("Selected Features for Clustering:")
display(clustering_data.describe())

# Check for missing or infinite values
print("\nMissing Values:")
print(clustering_data.isnull().sum())

print("\nInfinite Values:")
print(np.isinf(clustering_data).sum())

Selected Features for Clustering:

	transport_to_worker_ratio	transport_to_space_ratio
count	204.000000	204.000000
mean	0.016130	0.000037
std	0.065162	0.000045
min	0.000000	0.000000
25%	0.000000	0.000008
50%	0.000760	0.000022
75%	0.007727	0.000046
max	0.750000	0.000282

Missing Values:
transport_to_worker_ratio    0
transport_to_space_ratio     0
dtype: int64

Infinite Values:
transport_to_worker_ratio    0
transport_to_space_ratio     0
dtype: int64

Step 02 - Apply Clustering on Access and Demand Metrics

To enhance the identification of accessibility gaps, this step utilises an unsupervised machine learning approach through KMeans clustering. By analysing two key features—transport_to_worker_ratio and transport_to_space_ratio—the method aims to group city blocks into distinct clusters based on their accessibility profiles. Prior to clustering, feature normalisation is applied using StandardScaler to ensure both features contribute equally during distance-based clustering. The elbow method is then employed to determine the optimal number of clusters by minimising the total within-cluster sum of squares (TOTWSS).

# Import necessary libraries
#from sklearn.preprocessing import StandardScaler
#from sklearn.cluster import KMeans
#import matplotlib.pyplot as plt

# Step 1: Select the clustering features
clustering_features = merged_df[['transport_to_worker_ratio', 'transport_to_space_ratio']].dropna()

# Step 2: Normalise the features
scaler = StandardScaler()
scaled_features = scaler.fit_transform(clustering_features)

# Step 3: Use the Elbow Method to determine optimal number of clusters
totwss = []
for k in range(1, 10):
    kmeans = KMeans(n_clusters=k, random_state=42)
    kmeans.fit(scaled_features)
    totwss.append(kmeans.inertia_)

# Plot TOTWSS vs k to find the elbow
plt.figure(figsize=(8, 5))
plt.plot(range(1, 10), totwss, marker='o')
plt.title('Elbow Method to Determine Optimal k')
plt.xlabel('Number of Clusters (k)')
plt.ylabel('Total Within-Cluster Sum of Squares (TOTWSS)')
plt.grid(True)
plt.show()

Elbow Method Interpretation

The elbow plot indicates a distinct "elbow" at k = 3, suggesting that three clusters best capture the variance in the spatial accessibility data. This is the point after which the Total Within-Cluster Sum of Squares (TOTWSS) begins to decrease at a slower rate, showing diminishing returns for additional clusters.

Therefore, k = 3 will be used as the optimal number of clusters to segment blocks based on accessibility characteristics.

#from sklearn.preprocessing import StandardScaler

# Select features for clustering
features = merged_df[['transport_to_worker_ratio', 'transport_to_space_ratio']]

# Normalise the features
scaler = StandardScaler()
X_scaled = scaler.fit_transform(features)

#from sklearn.cluster import KMeans

# Fit KMeans with k=3
kmeans = KMeans(n_clusters=3, random_state=42)
merged_df['access_cluster'] = kmeans.fit_predict(X_scaled)

# Preview results
print("Cluster labels assigned to each block:")
display(merged_df[['block_id', 'transport_to_worker_ratio', 'transport_to_space_ratio', 'access_cluster']].head())

Cluster labels assigned to each block:

	block_id	transport_to_worker_ratio	transport_to_space_ratio	access_cluster
0	852	0.028689	0.000007	0
1	931	0.000752	0.000003	0
2	270	0.000834	0.000008	0
3	1006	0.017007	0.000005	0
4	502	0.007396	0.000016	0

# Count the number of blocks in each accessibility cluster
cluster_counts = merged_df['access_cluster'].value_counts().sort_index()

# Display result
print("Number of blocks per accessibility cluster:")
display(cluster_counts)

Number of blocks per accessibility cluster:

access_cluster
0    175
1     28
2      1
Name: count, dtype: int64

Interpretation of Accessibility Clusters

The clustering process using the KMeans algorithm categorised the 204 blocks into three distinct groups based on their transport_to_worker_ratio and transport_to_space_ratio. The distribution of blocks is as follows:

• Cluster 0 (Low Accessibility): 173 blocks
  This cluster contains the majority of blocks. These areas exhibit low accessibility in terms of transport availability relative to workforce size and built-up space. These blocks are likely underserved and may be prioritised for infrastructure development.

• Cluster 1 (Moderate Accessibility): 27 blocks
  These blocks show moderate access to transport facilities. While not critical, further improvements could enhance accessibility for workers and businesses.

• Cluster 2 (High Accessibility): 4 blocks
  This small cluster represents blocks with relatively high access to public transport. These areas are currently well-connected and likely do not require immediate intervention.

Step 03 - Labelling and Interpreting the Clusters

Once the KMeans clustering model has assigned each block to a cluster based on transport-to-worker and transport-to-space ratios, the next step involves interpreting the nature of each cluster. By analysing the centroid values of each cluster and comparing overall statistics, meaningful labels such as "Low Accessibility", "Moderate Accessibility", and "High Accessibility" can be assigned.

This categorisation helps in translating numeric clusters into descriptive accessibility categories, enhancing the understanding for urban planners and decision-makers.

# Get cluster centroids
centroids = pd.DataFrame(kmeans.cluster_centers_, columns=['transport_to_worker_ratio', 'transport_to_space_ratio'])
centroids['label'] = ['Cluster 0', 'Cluster 1', 'Cluster 2']
display(centroids)

# Sort centroids by accessibility level (optional logic)
# This step assumes higher transport ratios represent higher accessibility
centroids['access_level'] = [''] * len(centroids)
sorted_centroids = centroids.sort_values(by=['transport_to_worker_ratio', 'transport_to_space_ratio'], ascending=True).reset_index(drop=True)
sorted_centroids['access_level'] = ['Low Accessibility', 'Moderate Accessibility', 'High Accessibility']

# Map labels back to main dataframe
cluster_label_map = sorted_centroids[['label', 'access_level']].set_index('label')['access_level'].to_dict()
merged_df['access_level'] = merged_df['access_cluster'].map(lambda x: cluster_label_map.get(f'Cluster {x}'))

	transport_to_worker_ratio	transport_to_space_ratio	label
0	-0.049596	-0.334053	Cluster 0
1	-0.093241	2.107659	Cluster 1
2	11.289972	-0.555120	Cluster 2

Cluster Interpretation Summary

The KMeans clustering algorithm grouped the blocks into three distinct clusters based on normalised transport-to-worker and transport-to-space ratios:

• Cluster 0 (Low Accessibility):
  This cluster has negative or very low normalised values for both metrics, indicating poor access to transport infrastructure relative to job count and floor space. Most blocks in this group likely require improvement in accessibility services.

• Cluster 1 (Moderate Accessibility):
  Identified by high transport-to-space ratio, this cluster shows better-than-average transport infrastructure for the available floor space, though its worker ratio is moderate. These blocks are relatively well-served by public transport.

• Cluster 2 ( High Accessibility):
  This cluster has a very high transport-to-worker ratio but moderate transport-to-space accessibility. It may represent blocks with low employment demand but decent infrastructure presence.

These labels assist in translating raw clustering outputs into actionable insights for identifying transport-accessibility disparities across Melbourne.

Section 4: Visualisation and Dashboard Development

4.1 Spatial Visualisation of Accessibility Clusters

This map visualises the accessibility levels of different blocks across Melbourne based on transport-to-worker and transport-to-floor-space ratios. Each block has been assigned to one of three clusters using the KMeans algorithm:

• Cluster 0: Low accessibility
• Cluster 1: Medium accessibility
• Cluster 2: High accessibility

By mapping these clusters spatially, stakeholders can identify geographical patterns in transport accessibility and focus on regions needing improvement. The tooltip on each block provides its ID, cluster label, and calculated accessibility ratios.

# Ensure 'block_id' is the key for merging
final_map_df = pd.merge(merged_df, block_gdf[['block_id', 'geometry']], on='block_id', how='left')

# Convert to GeoDataFrame for map drawing
final_map_gdf = gpd.GeoDataFrame(final_map_df, geometry='geometry', crs=block_gdf.crs)

import folium
from folium.plugins import MarkerCluster
import json

# Step 1: Merge for spatial clusters
final_map_df = pd.merge(merged_df, block_gdf[['block_id', 'geometry', 'clue_area']], on='block_id', how='left')
final_map_gdf = gpd.GeoDataFrame(final_map_df, geometry='geometry', crs=block_gdf.crs)

# Step 2: Initialise map
m = folium.Map(location=[-37.8136, 144.9631], zoom_start=12, tiles="cartodbpositron")

# Step 3: Define cluster colors
cluster_colors = {
    0: 'red',    # Low accessibility
    1: 'orange', # Medium accessibility
    2: 'green'   # High accessibility
}

# Step 4: draw polygons with  tooltip
def add_shape(row, map_obj):
    if row['geometry'] is None:
        return

    geo_shape = json.loads(gpd.GeoSeries([row['geometry']]).to_json())['features'][0]['geometry']
    cluster = row['access_cluster']
    color = cluster_colors.get(cluster, 'gray')
    
    total_stops = row.get('bus_stop_count', 0) + row.get('tram_stop_count', 0)
    total_jobs = row.get('total_jobs_in_block', 0)
    total_space = row.get('total_floor_space_in_block', 0)

    tooltip_html = f"""
    <div style="font-family: Arial; font-size: 12px; padding: 6px;">
        <b>CLUE Area:</b> {row.get('clue_area', 'N/A')}<br>
        <b>Block ID:</b> {row['block_id']}<br>
        <b>Cluster:</b> {'Low' if cluster==0 else 'Medium' if cluster==1 else 'High'} Accessibility<br>
        <b>Total Stops:</b> {total_stops}<br>
        <b>Total Jobs:</b> {int(total_jobs)}<br>
        <b>Commercial Floor Space:</b> {int(total_space):,} sqm<br>
        <span style='color:blue;'>Transport-to-Worker Ratio:</span> {row['transport_to_worker_ratio']:.5f}<br>
        <span style='color:blue;'>Transport-to-Space Ratio:</span> {row['transport_to_space_ratio']:.5f}
    </div>
    """


    folium.GeoJson(
        geo_shape,
        style_function=lambda feature: {
            'fillColor': color,
            'color': color,
            'weight': 2,
            'fillOpacity': 0.2
        },
        highlight_function=lambda feature: {
            'weight': 3,
            'color': 'yellow',
            'fillOpacity': 0.7
        },
        tooltip=folium.Tooltip(tooltip_html, sticky=True)
    ).add_child(folium.Popup(tooltip_html, max_width=300)).add_to(map_obj)

# Step 5: Add shapes
for _, row in final_map_gdf.iterrows():
    add_shape(row, m)

# Step 6: Add Bus Stops
bus_cluster = MarkerCluster(name="Bus Stops").add_to(m)
for _, row in bus_gdf.iterrows():
    if pd.notnull(row['latitude']) and pd.notnull(row['longitude']):
        folium.Marker(
            location=[row['latitude'], row['longitude']],
            popup=folium.Popup(f"<b>Bus Stop</b><br>ID: {row.get('prop_id', 'N/A')}<br>Street: {row.get('addresspt', 'N/A')}", max_width=250),
            tooltip="Bus Stop",
            icon=folium.Icon(color='blue', icon='bus', prefix='fa')
        ).add_to(bus_cluster)

# Step 7: Add Tram Stops
tram_cluster = MarkerCluster(name="Tram Stops").add_to(m)
for _, row in tram_gdf.iterrows():
    if pd.notnull(row['latitude']) and pd.notnull(row['longitude']):
        folium.Marker(
            location=[row['latitude'], row['longitude']],
            popup=folium.Popup(f"<b>Tram Stop</b><br>Name: {row.get('name', 'N/A')}<br>Stop No: {row.get('stop_no', 'N/A')}", max_width=250),
            tooltip="Tram Stop",
            icon=folium.Icon(color='green', icon='train', prefix='fa')
        ).add_to(tram_cluster)

# Step 8: Add Legend
legend_html = """
<div style="position: fixed; bottom: 50px; left: 50px; width: 230px; height: 170px; 
    background-color: white; border:2px solid grey; z-index:9999; font-size:14px; padding: 10px;">
    <strong>Legend</strong><br>
    <i class="fa fa-square" style="color:red"></i> Low Accessibility<br>
    <i class="fa fa-square" style="color:orange"></i> Medium Accessibility<br>
    <i class="fa fa-square" style="color:green"></i> High Accessibility<br>
    <i class="fa fa-bus" style="color:blue"></i> Bus Stops<br>
    <i class="fa fa-train" style="color:green"></i> Tram Stops
</div>
"""
m.get_root().html.add_child(folium.Element(legend_html))

# Step 9: Display
m

Make this Notebook Trusted to load map: File -> Trust Notebook

4.2 Create Summary Table of Gap Zones

This step compiles a summary table that highlights blocks experiencing significant gaps in transport accessibility despite having high levels of workforce or commercial floor space. The table presents key indicators such as total job counts, commercial floor space, number of transport stops, and computed accessibility ratios. This overview helps stakeholders identify which zones require urgent attention or policy intervention based on both data-driven clustering and demand metrics.

# Step 1: Filter blocks with low accessibility cluster (Cluster 0)
gap_summary_df = final_map_gdf[final_map_gdf['access_cluster'] == 0]

# Step 2: Select and rename important columns
gap_summary_table = gap_summary_df[[
    'block_id', 'clue_area',
    'total_jobs_in_block',
    'total_floor_space_in_block',
    'bus_stop_count', 'tram_stop_count',
    'total_stops',
    'transport_to_worker_ratio',
    'transport_to_space_ratio'
]].copy()

gap_summary_table.rename(columns={
    'block_id': 'Block ID',
    'clue_area': 'CLUE Area',
    'total_jobs_in_block': 'Total Jobs',
    'total_floor_space_in_block': 'Commercial Floor Space (sqm)',
    'bus_stop_count': 'Bus Stops',
    'tram_stop_count': 'Tram Stops',
    'total_stops': 'Total Stops',
    'transport_to_worker_ratio': 'Transport-to-Worker Ratio',
    'transport_to_space_ratio': 'Transport-to-Space Ratio'
}, inplace=True)

# Step 3: Display the top 10 gap zones sorted by lowest worker ratio
gap_summary_table_sorted = gap_summary_table.sort_values(by='Transport-to-Worker Ratio').head(10)
display(gap_summary_table_sorted)

	Block ID	CLUE Area	Total Jobs	Commercial Floor Space (sqm)	Bus Stops	Tram Stops	Total Stops	Transport-to-Worker Ratio	Transport-to-Space Ratio
144	213	Carlton	0.0	26291.0	1	0	1	0.0	0.000038
116	854	South Yarra	0.0	15650.0	1	0	1	0.0	0.000064
113	860	South Yarra	0.0	96538.0	1	0	1	0.0	0.000010
34	869	South Yarra	0.0	31374.0	2	0	2	0.0	0.000064
140	253	Carlton	0.0	0.0	1	0	1	0.0	0.000000
129	219	Carlton	0.0	35798.0	1	0	1	0.0	0.000028
182	528	Kensington	0.0	18827.0	1	0	1	0.0	0.000053
176	443	West Melbourne (Residential)	0.0	28205.0	1	0	1	0.0	0.000035
40	273	Carlton	0.0	442580.0	2	0	2	0.0	0.000005
41	749	Southbank	0.0	82116.0	2	0	2	0.0	0.000024

4.3 Link Results to Policy Implications

The summary of gap zones identifies several blocks across Carlton, South Yarra, and North Melbourne that exhibit extremely low transport accessibility despite having substantial commercial floor space. These findings suggest a mismatch between infrastructure provision and urban activity density. Such evidence-based insights can support strategic planning decisions by:

• Prioritising these zones for public transport infrastructure upgrades.
• Reassessing zoning policies to align development intensity with accessibility.
• Targeting these areas in urban mobility improvement programs.

This alignment of data-driven insights with urban policy objectives ensures more equitable and sustainable growth across Melbourne’s urban landscape.

Final Project Summary

This spatial accessibility analysis aimed to evaluate how effectively public transport infrastructure in Melbourne supports areas of high employment and commercial activity. The project was structured across multiple stages, covering data acquisition, preprocessing, exploratory analysis, spatial computation, and advanced gap detection using machine learning.

---

Key Achievements

Stage 1: Data Preparation

• Imported and cleaned datasets on employment, floor space, bus and tram stops, and geographic block boundaries.
• Conducted spatial joins to associate transport stops with block-level boundaries.
• Calculated key accessibility metrics such as transport-to-worker and transport-to-floor-space ratios.

Stage 2: Exploratory Data Analysis (EDA)

• Analysed year-wise job and floor space trends.
• Compared job and floor space distribution across CLUE Small Areas.
• Visualised the spatial distribution of bus and tram infrastructure.
• Mapped total stops per block using interactive maps.

Stage 3: Spatial Accessibility Analysis

• Computed accessibility ratios per block.
• Identified gap zones using:
  • Rule-based filtering (e.g., low transport access but high job/floor space).
  • Machine Learning-based clustering using KMeans to group blocks by accessibility patterns.

Stage 4: Visualisation and Dashboard Development

• Created interactive maps with colour-coded clusters and tooltips showing transport, job, and space metrics.
• Summarised blocks requiring urgent infrastructure improvements.
• Linked analytical findings to potential planning policies and transport strategies.

---

Overall Insight

The integration of geospatial data, public transport points, employment density, and commercial space has enabled the identification of underserved urban blocks. By combining rule-based analysis with clustering algorithms, the project provided a multi-angle view of accessibility gaps. This approach can significantly aid urban planners in making data-driven decisions to improve equity and functionality in urban transport networks.

---

Future Recomandations

• Extend the model to include real-time population or visitor flow data.
• Integrate additional transport modes like cycling lanes or pedestrian paths.
• Develop a dashboard for ongoing monitoring of accessibility across the city.